found 1 place where panic with bridge+divert

[Alfred Perlstein]

I think i found one spot where having this setup causes a panic:
              ___________
 de0 (no IP) |FreeBSD box| xl0 - ips = 216.55.74.58, 192.168.2.1
             `-----------'

 i have bridge, and divert sockets along with ipfw in my kernel,
I'm having bridged packets filtered by ipfw via the sysctl.

about 1 second after enableing a divert rule on de0, i panic,
it seems to happen here:

/*      $Id: if_de.c,v 1.104 1999/05/03 09:36:29 dfr Exp $ */
"/usr/src/sys/pci/if_de.c" line 3739 of 5928 --63%-- col 28-31

at:

            nextout->d_addr1 = TULIP_KVATOPHYS(sc, mtod(ms, caddr_t));

for some reason i've been having a hell of a time trying to get gdb working
on this box and was stuck in DDB with an opcode: movl PTmap(,%eax,4),%eax

which is really in pmap.h (i386):

static __inline vm_offset_t
pmap_kextract(vm_offset_t va)
{
        vm_offset_t pa;
        if ((pa = (vm_offset_t) PTD[va >> PDRSHIFT]) & PG_PS) {
                pa = (pa & ~(NBPDR - 1)) | (va & (NBPDR - 1));
        } else {
                pa = *(vm_offset_t *)vtopte(va);
/* HERE */      pa = (pa & PG_FRAME) | (va & PAGE_MASK);
        }
        return pa;
}

people in -stable seem not to have problems with bridge+divert sockets,

anyone have any ideas?


btw, if i enable the divert socket on my 'xl' card then it also
seems to happen in an intrupt context.

btw, this newbus stuff looks sorta interesting, considering it took 
me about a week to understand the old way, is there any chance for
an API to be published?  any good pointers to know what i can do
with this stuff?

thanks,
-Alfred 



To Unsubscribe: send mail to majord...@freebsd.org
with "unsubscribe freebsd-current" in the body of the message